Modeling Methodology for Component Reuse and System Integration for Hurricane Loss Projecti

Modeling Methodology for Component Reuse and System Integration for Hurricane Loss Projection ApplicationKasturi Chatterjee1, Khalid Saleem1, Na Zhao1, Min Chen1, Shu-Ching Chen1, Shahid S. Hamid21Distributed Multimedia Information System LaboratorySchool of Computing and Information SciencesFlorida International University, Miami, FL 33199, USA2Department of FinanceFlorida International University, Miami, FL 33199, USA1{kchat001, ksale002, nzhao002, mchen005, chens}@, 2hamids@AbstractHurricanes are one of the deadliest and perilous natural calamities on the face of earth having a severe impact both on the lives of the people and economy of a nation. Attempts have been made to mitigate hurricane after-math, by utilizing research and tools that can analyze hurricanes and estimate projected losses. The need for such research methodologies and tools stimulated the de-velopment of a multi-disciplinary cutting edge public hur-ricane model called Public Hurricane Risk and Insured Loss Projection Model (PHRLM). The complex and di-verse nature of the application raises the need for module abstraction, seamless integration and effective reusability to create a uniform generic environment. Providing effi-cient interaction between these complex multi-disciplinary modules using different abstractions, formal-isms, data formats and communications and making each module transparent enough to be reusable becomes a complicated task. This paper presents a UML based for-mal Modeling Methodology, enabling component reuse and integration of the application.1. IntroductionHurricanes pose one of the greatest environmental threats to human beings and economy of the nations. For the past few years, hurricanes have been more frequent in U.S due to increased warming of water in the Atlantic Ocean causing loss of human lives and destruction of properties in the coastal regions. Over the period of time, different commercial modeling companies have devel-oped risk and assessment models to assist insurance com-panies in formulating their rating policies. These models lack transparency in describing the assumptions and ra-tionales used to estimate the losses and leave the general public and regulators at the whims of the companies, to accept the scenarios as presented without being able to judge their credentials and accuracy. The public systems like HAZUS [6], though have the transparency of infor-mation lack portability and easy maintenance. Due to their PC-based implementation, new hardware needs to be deployed with the increase of users and installations need to be repeated. Moreover, the different modules of the application execute rather independently making informa-tion exchange, reuse and efficient resource management a hurdle. In order to address all the above stated problems, a multi-disciplinary public model was developed called Public Hurricane Risk and insured Loss projection Model (PHRLM) [9] by the State of Florida, funded by Florida Office of Insurance Regulation. The model has a compo-nent-based approach where each sub-application corre-sponding to a different genre has been modularized. The model was developed by an interdisciplinary team of ex-perts belonging to meteorology, statistics, finance, actuar-ial science, engineering and computer science disci-plines. It is an open model facilitating ease of maintainability and portability. The complexity and diver-sity of the system makes system integration a challenge. Different modules carrying domain specific tasks need to seamlessly interact with one another which necessitate semantic maintenance and data integrity as well as the presence of efficient interfaces enabling abstraction. Moreover, to enhance the utility of the application, the individual components should be reusable by different applications with minimum system customization.The diverse and complicated structure of the PHRLM ap-plication necessitates the introduction of a modeling tech-nique to achieve the desired abstraction required both for efficient integration as well as independent component re-use. A model is an abstract representation of a system that enables us to answer questions about the system [2]. Mod-eling is a way to deal with complexity by ignoring irrele-vant details. Efficient system integration and reuse requires an object-oriented and component based approach,Figure 1. System Architecture of PHRLMas the main essence of object oriented paradigm is reuse and encapsulation. Hence, to model the application, UML was found to be the appropriate language as the main goal of UML is to provide a standard notation that can be used by all object-oriented methods and to select and integrate the best elements of precursor notations [2]. Model driven reuse and integration differs from integra-tion and reuse by programming. Programming approach concentrates on a particular scenario and inextensible so-lution to a specific challenge. Model-driven integration focuses on abstracting the information content into a model that describes the application's information re-sources and thus makes it easy to be applicable to any system within the same genre. This paper proposes a UML based modeling technique to enable and enhance efficient system integration and component reuse for the PHRLM application and serves as a prototype to solve similar problems for multi-disciplinary large-scale research applications. The rest of the paper is organized as follows. Section 2 presents a brief overview and System Architecture of the PHRLM Project, Section 3 discusses the Modeling meth-odologies undertaken to successfully formalize the appli-cation with a focus to enable component reuse and system integration modeling techniques followed by Section 4 which presents the conclusion. 2. Overview of PHRLMThe PHRLM model is a probabilistic model designedto estimate damage and insured losses due to the occur-rence of hurricanes in the Atlantic Basin. The PHRLMestimates the full probabilistic distribution of damage andloss for any significant storm event. The modeling meth-odology of PHRLM can be partitioned into three major components: (a) Wind Module, (b) Vulnerability Module and (c) Insured Loss Module. The major components are developed independently before integration. The Wind Hazard Module is a meteorological module which deals with estimating the number of hurricanes and storm gene-sis time, generating storm track, open terrain wind speeds and ultimately the wind probability for each zip code. Thenext module is the Vulnerability Module which is an en-gineering module which simulates the wind damages, generates the damage matrices and the vulnerability func-tions for mitigated structures. The final module is the In-sured Loss Module which models the wind deductibles, generates the annual loss and the expected loss cost for a specific hurricane. The model can estimate losses in a particular zip code and can also simulate a potential storm. Each module can be considered as a business ob-ject component which provides a service-oriented inter-face to external systems and allows interoperability be-tween systems.2.1. System ArchitectureAn appropriate system architecture provides flexible yet robust infrastructures to build, extend and maintain any application. The PHRLM system uses a modular compo-nent-based architecture. PHRLM being a multi-disciplinary project, different modules relate to one or more domains. Due to the research nature of this applica-tion, each module is subject to constant revision and up-date. Thus, the modules are made as loosely coupled as possible to meet the flexibility and adaptability require-ments. In Figure 1, the system architecture of the applica-tion is illustrated.Figure 2. Use Case Diagram of PHRLM3. Modeling MethodologyThe multi-disciplinary nature of the project makes inte-gration and component reuse a challenging job as it in-creases the complexity of the system manifold. Thus, arose the need of modeling to provide the necessary ab-straction required to achieve seamless integration and re-use efficiently. The modular nature of the system and the reusability requirement makes object-oriented modeling apt for the job.3.1. Enabling Component Reuse Using UMLWith the increase of complexity and size of applications, development costs have increased manifold. Thus, effi-cient reuse of existing technologies has become very im-portant. Object-oriented paradigm enables easy reuse of components. An object-oriented framework is defined as a set of classes that embodies an abstract design for solu-tions to a family of related problems [2].Since, our application is a public-benefiting open model; one of the main requirements was reuse by other similar applications. As explained earlier, there are several mod-ules in the system performing specific tasks. For example, the Vulnerability Module calculates the damage ratios with the supplied values of vulnerability statistics and wind speed. The same module has the potential to be ex-tended for being reused in calculating the damage caused by other kind of wind based natural calamity like Torna-dos etc. by appropriate control of the input parameters and customization. Moreover, currently the model can es-timate hurricane losses for only residential structures. However, the model is highly extensible and allows for the addition of components to estimate losses for com-mercial structures and high rise buildings. The model can also be used as a reference to develop similar models for other vulnerable coastal areas and can be extended to de-velop general Disaster Control and Management models. For effective reuse several criteria should be satisfied by each module as well as by the entire system. The module should have a proper abstraction to make the complex functionalities transparent to the user, it should be port-able and should be able to exist and co-operate independ-ently outside the existing framework. Efficient compo-nent reuse also needs the entire system to be loosely coupled so that the dependence of each module with one another is as minimal as possible. To fulfill all the above requirements, a modeling technique is required. UML provides the required abstraction and a complete stan-dardized description of the system. Each UML diagram is designed to let developers and customers view a software system from a different perspective and varying degree of abstraction. UML diagrams commonly include Use Case Diagrams, Class Diagrams, Interaction Diagrams, State Diagrams, Activity Diagrams and physical Diagrams. Major reuse scenarios can be modeled by three generali-zation relationships supported by Use Case Diagrams [5] viz. <extend>, <include> and <inheritance>. An extend-ing use case is an alternate course of the base use case. An include dependency is a generalization relationship denoting the inclusion of the behavior described by an-other use case. The third way of reuse is inheritance of use cases where one use case is inherited from other use cases and the inheriting use case would completely re-place one or more of the courses of action of the inherited use case. The introduction of these generalization tech-niques allows reuse both within the application as well as by other external applications.Based on the component reuse frameworks proposed in [7, 8], two major characteristics need to be considered: class generality assessment and relations among classes. As proposed in [7, 8], a class may be marked as general or specific . A general class is expected to be reused but a specific class is intended for a particular application. This concept is extended and used in modeling of our system. Since every component is aimed to be reusable, hence classes belonging to different modules should be a gen-eral class. However within a particular module, the class can be specific . Figure 2 denotes the Use Case diagram of the entire system which depicts the implementation of <include> relationship among different modules. To maintain the generality of the classes and to make each module independent, <extend> or <inheritance> rela-tionship is not used within the system. Thus, the individ-ual modules can be easily extended or inherited in other applications without any constraint of inter-modular de-pendability. Figure 3 explains one such scenario where the Vulnerability Module could be extended to different types of application specific estimation scenarios like theDamage Estimation for other Wind Associated Natural Calamity and the generation of Damage Matrices and Vulnerability Function for Commercial buildings. The modeling technique applied to PHRLM aiming at compo-nent reuse can be formalized as:Definition 1:If, there is a loosely coupled system with self sufficient in-teracting modules, each module will be modeled using a separate Use Case.Definition 2:Suppose there exist two Use Cases UC A and UC B for modules A and B respectively. The only generalization relation that can exist between A and B is UC A <include> UC B due to the loosely coupled nature of the system. However, in order to reuse a generic Use Case UC A,it should have a generalization relationship of <extend> with any customized external Use Case.Once the Use Cases have been defined, the class dia-grams are required to model classes. Class Diagrams de-scribe the structure of the system in terms of classes and objects. The reuse potential of class depends on the extent to which class to class relationship or coupling is defined [7, 8]. Classes are related to other classes if they expect to use and reuse other classes in present implementations and future applications [5]. The class to class relationship is very useful to plug in a component from one applica-tion to another. One of the ways to design the class rela-tionship is by formulating the use case to a class relation-ship, i.e., by keeping track of the classes included in each use case and then using the use case relationship to de-termine if the classes are related. Figure 4 denotes the Class Diagram of a component of the PHRLM model which explains the class relationships and hierarchy in detail. For example, the class fitDistriBean is related to SimuSelection. Hence in order to reuse fitDistriBean in some other application, the class SimuSelection needs to be present in that particular application too. Thus, if a component of an application was to be reused, the de-tailed class relationship gives the flexibility of functional application-nature dependent modification and customiza-tion. The rationale used behind modeling the classes of PHRLM to help efficient component reuse is as follows: Definition 3:Within the classes of the same Use Case, <extend> rela-tionship can exist but not between classes of different Use Cases.The three definitions described above form the basic ra-tionale by which efficient component reuse was modeled for a diverse loosely coupled system where modules communicate with each other using Data Flow. The ap-proach provides a modeling prototype for any Disaster Management Application consisting of several compo-nents performing different domain specific jobs. Thus, with proper UML modeling using Use Case Diagrams and Class Diagrams, the different components of the PHRLM system can be customized and plugged in by other applications efficiently.Figure 4. Modeling of Class level Coupling of aPHRLM Module3.2. Enabling Component Integration Using UML The multi-disciplinary nature of PHRLM makes system integration an indispensable but a very complex task. Maintaining semantic uniformity of the data as well as their seamless communication among the different mod-ules becomes a challenging job. Due to the diverse nature of the project, the integration of PHRLM could not be achieved by following any one of the popular integrating techniques like File Transfer, Shared Database, RemoteProcedure Invocation or Messaging in their absolute forms, instead it required a combination of these tech-niques and customization at some places to suit the re-quirements of the project. From the implementation point of view, semantic integration is achieved by selecting the key terms across the entire application and making them uniform. Due to domain specific requirements, these key terms have different local meanings which were resolved using a mapping technique that involved the mapping of local terms with the global terms. Asynchronous messag-ing technique is used to integrate the loosely coupled components of the system by overcoming the limitations of latency and unreliability. To achieve the above tech-nique, each module is developed into a web-based appli-cation and they are made to communicate through Event Messaging and Document Messaging over a secured channel. For data that is too voluminous or too sensitive to be transmitted over communication channel, a central-ized database is used for storage, maintenance and effi-cient access of sensitive data.Minimum Pressure Central PressureFigure 5. Semantic Integration Modeling for Global Concept “Minimum Pressure”The above integration methodologies and planning for PHRLM explicitly depicts the use of several integration approaches which calls for an efficient modeling repre-sentation of the entire integration approach for formaliza-tion and efficient understanding. The integration method-ologies adopted in PHRLM cannot be compared with any Enterprise Integration Architecture as it has an essential research approach which calls for constant revision and update of requirements as well as design thus hindering it from following the strict life-cycle of any enterprise inte-gration architecture (e.g. CIMOSA [3] and ARIS [1] ) which is composed of domain identification, concept de-sign, requirement definitions, design specifications, im-plementation description, domain operation and decom-mission definition [10]. Here, we propose an integration modeling specification for PHRLM which can be ex-tended for other Disaster Management and Recovery Sys-tems. The modeling involves combination of the UML Data Modeling Technique [4] and Use Case and Class Diagrams to depict the integration plans utilized by PHRLM. In our system, Semantic Integration is modeled using Use Case Diagrams with generalization relationship de-picting the semantic uniformity. For each global term allthe modules using it are related to one another by a <in-clude> relationship to depict the flow. The method can be formalized with the following definition.Definition 4:If Module A, Module B and Module C use a global con-cept I with different local terms, then there exist a Use Case U I for the global concept I which consists of the use cases for Module A, B and C and the local terms in each of them related to each other by a <include> relationship depicting the presence of the Global Term I in each one of them.The above definition is explained in Figure 5 where the use of the global concept “the minimum pressure of the hurricane” in different modules is illustrated. In Wind Field Module, the term Minimum Pressure is synonymous to the term Central Pressure in Storm Track module, both depicting the concept of minimum pressure of a hurri-cane.The integration through messaging is modeled using Se-quence Diagrams for pairs of modules that communicate with one another. The sequence diagram for individual module also depicts the communication among the differ-ent sub-modules within the module. The proposed con-cept is conceptualized with the following definition.Figure 6. Asynchronous CommunicationModeling across ModulesDefinition 5:If Module A communicates with Module B, then there ex-ists a Sequence Diagram S AB which depicts the communi-cation sequence among them.System Integration using Centralized Database is mod-eled using the UML Data Modeling Profile as introduced in [4]. Here, the concept of tables and relationships used in a database maps to the concept of classes and associa-tions in UML. The database used in PHRLM has several components like Tables, Schemas, TableSpace, View, Columns, Key, Index and Constraints. Each of them is modeled using the UML Data Modeling Technique. Some of them are explained below:1.Tables: A table represents a set of records having asimilar structure containing data. A table is repre-sented in the UML diagram as class diagrams.2.Schemas: A Schema is used to organize tables. InUML modeling, a schema is represented by a pack-age.3.Columns: Column is the field in the table where datais stored. Columns are modeled using Attribute rep-resentation in UML.4.Constraints: Constraint is a rule imposed on the col-umns or on rows of a database. They are modeled through stereotyped operations as well.The data modeling techniques discussed are illustrated in Figure 7. The detailed modeling of the database using easy to understand UML diagrams helps in integration and use of the centralized database by all the participating modules. Such detailed representation helps in easy communication and adaptation of any change in the cen-tralized database. This leads to a uniform and detailed un-derstanding of the database schema by all the participat-ing modules, thus facilitating the integration process without suffering from inconsistency and ambiguous in-terpretation.Thus, we propose a unique modeling technique for inte-gration of complex, multi-disciplined and loosely coupled systems by combining three types of modeling ap-proaches applied to semantic integration, communication and centralized databases. The approach can be easily ex-tended and used as a prototype for developing and inte-grating other similar applications.4.ConclusionThis paper proposes a modeling methodology for com-ponent reuse and system integration for a diverse multi-disciplinary application PHRLM, which analyzes hurri-canes to predict their paths and occurrence and estimate projected losses. The proposed modeling methodology is unique and customized for a loosely coupled, multi-domain large scale research oriented application and can be used as a prototype to design Disaster Management and Recovery Systems for different kinds of natural ca-lamities. It will also assist in solving imperative problems of system integration, extensibility and component reuse which are a critical concern for successful software engi-neering implementation.5. AcknowledgementThis work is partially supported by Florida Office of In-surance Regulation (OIR) under “Hurricane Loss Projec-tion Model.” While the project is funded by the Florida Office of Insurance Regulation, the OIR is not responsi-ble for this paper content.References[1]ARIS Reference, “http://www.ids-/international/english/products/aris_design_platform/50324”.[2] B. Bruegge and A.H. Dutoit, “Object-oriented SoftwareEngineering Using UML, Patterns, and Java,” Second Edi-tion, 2004.[3]CIMOSA Reference, “t.pl”.[4]Davor Gornik, “UML Data Modeling Profile”, White Pa-per, Rational Software. May 2002.[5] D. Needham, R. Caballero, S. Demurjian, F, Eickhoff, J.Mehta and Y. Zhang, “A Reuse Definition, Assessment,and Analysis Framework for UML”, Book Chapter in Ad-vances in UML and XML –Based Software Evolution,2005.[6]Hazus manuals page,“/hazus/li_manuals.shtm”.[7]M. Price and S.A. Demurjian, “Analyzing and MeasuringReusability in Object-Oriented Design,” Proceedings ofOOPSLA’97, 1997.[8]M. Price, S. Demurjian and D. Needham, “ReusabilityMeasurement Framework and tool for Ada95,” Proceed-ings of 1997 TriAda Conf., Nov.1997.[9]PHRLM manual,“/hurricaneloss”.[10]Y. Zhou, Y. Chen and H. Lu, “UML-based Systems Inte-gration Modeling Technique for the Design and Develop-ment of Intelligent Transportation Management System,”Proceedings of IEEE International Conference on Systems, Man and Cybernatics, 2004.。